Upload
others
View
12
Download
0
Embed Size (px)
Citation preview
IAENG International Journal of Computer Science, 35:1, IJCS_35_1_03
Image Encryption Using Block-BasedTransformation Algorithm
Mohammad Ali Bani Younes and Arnan Jantan
407
(Advance online publication: 19 February 2008)
Abstract-Encryption is used to securely transmit data inopen networks. Each type of data has its own features thereforedifferent techniques should be used to protect confidentialimage data from unauthorized access. Most of the availableencryption algorithms are mainly used for textual data and maynot be suitable for multimedia data such as images. In thispaper, we introduce a block-based transformation algorithmbased on the combination of image transformation and a wellknown encryption and decryption algorithm called Blowfish.The original image was divided into blocks, which wererearranged into a transformed image using a transformationalgorithm presented here, and then the transformed image wasencrypted using the Blowfish algorithm. The results showedthat the correlation between image elements was significantlydecreased by using the proposed technique. The results alsoshow that increasing the number of blocks by using smallerblock sizes resulted in a lower correlation and higher entropy.
Index Terms-Image correlation, Image encryption, Imageentropy, Permutation.
I. INTROD UCTION
The rapid growth of computer networks allowed large files ,such as digital images, to be easily transmitted over theinternet [I]. Data encryption is widely used to ensure securityhowever, most of the available encrypti on algorithms areused for text data. Due to large data size and real timeconstrains, algorithms that are good for textual data may notbe suitable for multimedia data [2]-[4].
In most of the natural images, the values of theneighboring pixels are strongly correlated (i.e. the value ofany given pixel can be reasonabl y predicted from the valuesof its neighbors) [5]-[7] . In order to dissipate the highcorrelation among pixels and increase the entropy value, wepropose a transformation algorithm that divides the imageinto blocks and then shuffles their positions before it passesthem to the Blowfi sh encryption algorithm. By using thecorrelation and entropy as a measure of security, this processresults in a lower correlation and a higher entropy value whencompared to using the Blowfish algorithm alone, and thusimproving the security level of the encrypted images. Thereare two main keys to increase the entropy; the variable secretkey of the transformation process and the variable secret keyof the Blowfish algorithm.
Manuscript received 25 lui 2007 . Mohammad Ali Moh'd Bani Younesand Aman Jantan are with the Department of Computer Science, UniversitiSains Malaysia. and E-mail: mali [email protected] @cs.usm.my.
The variable secret key of the transformation processdetermines the seed, which is used to build the secrettransformation table with a variable number of blocks. If thekey is changed, another seed will be generated, and then adifferent secret transformation table is obtained.
The variable secret key of the Blowfish algorithm is usedto encrypt the transformed image . This encrypt ion proces sdecrease s the mutual information among the encrypted imagevariables (i.e. high contrast) and thus increasing the entropyvalue. In this paper we propose a block-based transformationalgorithm in order to increase the securit y level of theencrypted images. The rest of this paper is organized asfollows. Section 11 gives a background about the currentimage encryption schemes. In Section III, the description ofthe proposed block-based transformation algorithm ispresented. Section IV and V present the experimental resultsand discussion, while section VI presents a comparisonbetween the proposed algorithm and those of other systems .Finally, section VII concludes the paper.
II. BACKGROUND
Encryption is the process of transforming the informationto insure its security. With the huge growth of computernetworks and the latest advances in digital technologies, ahuge amount of digital data is being exchanged over varioustypes of networks. It is often true that a large part of thisinformation is either confidential or private. As a result,different security techniques have been used to provide therequired protection [8].
The security of digital images has become more and moreimportant due to the rapid evolution of the Internet in thedigital world today. The security of digital images hasattracted more attention recently, and many different imageencryption methods have been proposed to enhance thesecurity of these images [9].
Image encryption techniques try to convert an image toanother one that is hard to understand [9]. On the other hand,image decryption retrieves the original image from theencrypted one. There are various image encryption systemsto encrypt and decrypt data, and there is no single encryptionalgorithm satisfies the different image types.
Most of the algorithms specifically designed to encryptdigital images are proposed in the mid-I 990s. There are twomajor groups of image encryption algorithms: (a) non-chaosselective methods and (b) Chaos-based selective ornon-selective methods. Most of these algorithms are
IAENG International Journal of Computer Science, 35:1, IJCS_35_1_03
408
(Advance online publication: 19 February 2008)
Fig. I. General block diagram of the transformationalgorithm
The transformation algorithm is presented below. It generatesa transformation table that will be used to build a newlytransformed image.
OriginalImage
EncryptionProcess
Transformationprocess
information can be reduced by decreasing the correlationamong the image elements using certain transformationtechniques.
The secret key of this approach is used to determine theseed . The seed plays a main role in building thetransformation table, which is then used to generate thetransformed image with different random number of blocksizes. The transformation process refers to the operation ofdividing and replacing an arrangement of the original image.
The image can be decomposed into blocks; each one
contains a specific number of pixels. The blocks aretransformed into new locations. For better transformationthe block size should be small, because fewer pixels keeptheir neighbors. In this case, the correlation will bedecreased and thus it becomes difficult to predict the valueof any given pixel from the values of its neighbors. At thereceiver side , the original image can be obtained by theinverse transformation of the blocks. A general blockdiagram of the transformation method is shown in Fig. I.
ALGORITHM CREATE_TRANSFORMATION_TABlEI : Load Image2: Input key
3: Get ImageWidth and ImageHeight4:
4.1 : LowerHorizontalNoBlocks = Int(ImageWidth /I 0)4.2: LowerVerticalNoBlocks = Int(lmageHeight /10)
5: Randomize 06:
6.1: HorizontalNoBlocks = RandomNum between(LowerHorizontalNoBlocks and ImageWidth)
6.2: VerticalNoBlocks = RandomNum between(LowerVerticalNoBlocks and ImageHeight)
7: NoBlocks = HorizontalNoBlocks * VerticalNoBlocks8: Seed = IHash value (Key) I9: HashValuel = [Hash value (first half of the Key)l
HashValue2 = [Hash value (second halfofthe Key)l10: Randomize using seed
II: IfHashValuel > HashValue2 Then
designed for a specific image format compressed oruncornpressed, and some of them are even format compliant.There are methods that offer light encryption (degradation),while others offer strong form of encryption. Some of thealgorithms are scalable and have different modes rangingfrom degradation to strong encryption [10].
Mitra A et al. [II] have proposed a random combinationalimage encryption approach with bit, pixel and blockpermutations.
Zhi-Hong Guan et al. [12] have presented a new imageencryption scheme, in which shuffling the positions andchanging the grey values of image pixels are combined toconfuse the relationship between the cipher image and theplain image.
Sinha A. and Singh K. [13] proposed an image encryptionby using Fractional Fourier Transform (FRFT) and JigSawTransform (JST) in image bit planes.
Shujun Li et al. [14] have pointed out that allpermutation-only image ciphers were insecure againstknown/chosen-plaintext attacks. In conclusion, theysuggested that secret permutations have to be combined withother encryption techniques to design highl y secured images.
Maniccam S.S. and Bourbakis N G. [10] proposed imageand video encryption using SCAN patterns. The imageencryption is performed by SCAN-based permutation ofpixels and a substitution rule which together form an iteratedproduct cipher.
Ozturk I. and Sogukpinar I. [15] proposed new schemeswhich add compression capability to the mirror-like imageencryption MIE and Visual Cryptography VC algorithms toimprove these algorithms.
Sinha A. and Singh K. [16] proposed a technique toencrypt an image for secure transmission using the digitalsignature of the image . Digital signatures enable the recipientof a message to authenticate the sender of a message andverify that the message is intact.
Droogenbroeck M.V. and Benedett R. [2] have proposedtwo methods for the encryption of an image; selectiveencryption and multiple selective encryption.
Maniccam S.S., Nikolaos G. and Bourbakis. [17] havepresented a new methodology, which performs both losslesscompression and encryption of binary and gray-scale images.The compression and encryption schemes are based onSCAN patterns generated by the SCAN methodology.
The proposed algorithm divides the image into randomnumber of blocks with predefined maximum and minimumnumber of pixels, resulting in a stronger encryption and adecreased correlation.
III. THE PROPOSED TECHNIQUE
A. Description ofthe Transformation Algorithm
The transformation technique works as follows: theoriginal image is divided into a random number of blocksthat are then shuffled within the image. The generated (ortrans~ormed) image is then fed to the Blowfish encryptionalgonthm. The main idea is that an image can be viewed as an~rrang~ment o.f blocks. The intelligible information presentIII an Jma~e IS dU~ to the correlation among the imageelements III a given arrangement. This perceivable
IAENG International Journal of Computer Science, 35:1, IJCS_35_1_03
SEEDALTERNATE = IElse
SEEDALTERNATE = 2End If
12: I = 0Number-of-seed-changes (N) = I
13: While I < NoBlocksR = RandomNum between (zero and NoBlocks -I)If R is not selected Then
Assign location R to the block I1+=1
ElseIfSEEDALTERNATE = I Then
seed = seed + (HashValuel Mod I) +1SEEDALTERNATE = 2Elseseed = seed + (HashValue2 Mod I) + ISEEDALTERNATE = I
Randomize (seed)End IfElse
Number-of-seed-changes += IIf Number-of-seed-changes > 500,000 then
For K = 0 to NoBlocks -IIf K not selected then
Assign location K to Block I1=1+I
End ifNext K
End ifEnd ifEnd WhileEND CREATE_TRANSFORMATION_TABLE
Input: BMP image file, a string KeyOutput: Transformation table
ALGORITHM PERFORM_TRANSFORMATIONI: For I = 0 to NoBlocks -I
1.1: Get the new location of block I from thetransformation table1.2: Set block I in its new locationEND PERFORM_TRANSFORMATION
Input: Original Image (BMP image file) andTransformation tableOutput: Transformed Image.
B. Description ofthe combination technique
The block-based transformation algorithm is based on thecombination of image transformation followed byencryption (i.e. transformation algorithm followed by theBlowfish algorithm). The transformation algorithm and theBlowfish algorithm use the original image to produce threeoutput images; (a) a ciphered image using Blowfish, (b) atransformed image using a transformation process and (c) atransformed image encrypted using Blowfish.
The correlation and entropy of the three images arecomputed and compared with each other. This techniqueaims at enhancing the security level of the encrypted images
by reducing the correlation among image elements andincreasing its entropy value.
Image measurements (correlation and entropy) will becarried out on the original image and the encrypted imageswith and without transformation algorithm the results arethen analyzed. The overview model of the proposedtechnique is shown in Fig.2.
Fig. 2. An overview diagram of the proposed technique
IV. EXPERIMENTS
The method used to evaluate the present technique isdescribed in Fig. 2. The algorithm was applied on a bitmapped (bmp) image that has the size of 300 pixels x 300pixels with 256 colors. In order to evaluate the impact of thenumber of blocks on the correlation and entropy, threedifferent cases were tested. The number of blocks and theblock sizes for each case are shown in Table I.
Table I Different Cases to Test the Impact of the Number ofBlocks on the Correlation and Entropy
Case Number Number of blocks Block size
I 301'30 10 pixels I' 10 pixels
2 60 x60 5 pixels I' 5 pixels
3 1001'100 3 pixels x 3 pixels
Each case produces three output images; (a) a cipheredimage using the Blowfish algorithm, (b) a transformed imageusing the proposed algorithm, and (c) a ciphered image usingthe proposed algorithm followed by the Blowfish algorithm.For the rest of this paper, we use image A, image B, image C,and image D to denote the original image, the ciphered imageusing the Blowfish algorithm, the transformed image, and theciphered image using the proposed algorithm followed by theBlowfish algorithm respectively.
Correlation and entropy are computed for each caseaccording to equation (I) and equation (2).
409
(Advance online publication: 19 February 2008)
IAENG International Journal of Computer Science, 35:1, IJCS_35_1_03
(c ) (d)Fig. 5. Results of encryption by using 5 pixels x 5 pixelsblocks. (a) Original image. (b) Encrypted image usingBlowfish. (c) Transformed image. (d) Encrypted imageusing transformed followed by the Blowfish algorithm.
The correlation and entropy results of this case aresummarized in Table III and Fig. 6.
(b)Fig. 4. Correlation and entropy values of Case I. (a)Correlation between directions. (b) Average correlation andentropy.
Case 2: The image is decomposed into 5 pixels x 5 pixelsblocks. Fig. 5 shows the resulted images.
(b)
Table 11 Results of Correlation and Entropy values of Case 1.
(c) (d)Fig. 3. Results of encryption by using 10 pixels x 10 pixelsblocks. (a) Original image. (b) Encrypted image usingBlowfish. (c) Transformed image. (d) Encrypted imageusing transformed followed by the Blowfish algorithm.
The correlation and entropy results of this case aresummarized in Table II and Fig. 4
Measurement A B C D
Horizontal 0.933 0.035 0.843 0.034
= Vertical 0.936 0.107 0.844 0.038.S'; Diagonal 0.919 0.032 0.748 0.013~t:
Opposite00.916 0.034u Diagonal
0.745 0.009
Average 0.926 0.052 0.795 0.023
Entropy value 2.431 4.799 2.431 5.231
nL CxY)-llU 1.1
r~nLC~) -CLx)2 InL C/ )-Cui]
(1 )0.9 -----Where c 0.7
r: correlation value.2~ 0.5
n: the number of pairs of data e5
I xy: sum of the products of paired data c 0.3
I x: sum of x data0.1
Iv: sum ofy data------
I .l: sum of squared x data -0.1
:Ll: sum of squared y data Horizontal Vertica l Diagonal OppositeDiagonal
Entropy defined as follows [18]-[19]. --+-A . B - -c - - 0
G - I
H e =-'L P Ck ) log 2 CP Ck )) ( 2) (a)
k =O
Where:1.1 5.5
He: entropy. 0.9· 5
G: gray value of input image (0... 255).P(k): is the probability of the occurrence of symbol k. c · 4.5
0 0.7 .~~ 4 »,
Case 1: The image is decompo sed into 10 pixels x 10 pixels 0c,
o 0.5 . g
blocks. Fig.3. shows the resulted images.Q>
cCl
· 3 .5 w~Q>
0.3>« 3
0.1 · 2.5
-0.1 · 2
A B C 0
410
IAENG International Journal of Computer Science, 35:1, IJCS_35_I_03
· 2
6
55
· 5
· 4 .5 ei:· 4 e
"E· 3 .5 W
· 3
2 .5
OppositeDiagona l
D•
Diagonal
c
•
~'i'rt>----A
Vertical
(a)
•
•
B
•
•
. A __ B --&-- C ->f--- D
•
A
•
Horizontal
1
0 .9
.~ 0 .8
12 0 .7 ··
e 0.6(;o 0 .5
~ 0.4 •co4; 0.3 '
~ 0 .2 ··
0 .1
o ..
1.1
0.9
c 0 .70
""'" 0.5~0o 0.3
0.1
-0.1
Measurement A B C D
Horizontal 0.9325 0.0346 0.6289 0.0129
Vertical 0.9362 0.1073 0.6265 0.0034c0
~Diagonal 0.9186 0.03 21 0.4 145 0.0014
~t:0 Opposite 0.9156 0.0337 0.4146 0.0049u
Diagonal
Aver age 0.9257 0.0519 0.5211 0.0056
Entropy value 2.4305 4.799 2.4305 5.5281
Oppos iteDiagonal
Diagonal
•
Vert ical
(a)
._-_..
--+- A ·-- ·- B __ C __ D
•
t=:=--------------------Horizontal
1 5
0 .9 • •c: 4 .5.Q 0 .8
12 0 .7 • <, 4e 0.6 / >-(;
o,o
0 0 .5 3 .5 -EQ) 0 .4en ; Wco ; 3... 0 .3Q)
0 .2~ .,."... . . . _ .. • 2.5
0.1 I •0 ~ -- -- . .... . 2
A B C D
1.1
0 .9
c 0.70
~ 0.5~8 0.3 -
0 .1
-0.1
(b)Fig. 6. Correlation and entropy values of Case 2. (a)Correlation between directions. (b) Average correlation andentropy.
Case 3: The image is decomposed into 3 pixels x 3 pixelsblocks. Fig. 7 shows the resulted images.
Table III Results ofCorrela tion and Entropy values ofCase 2. Fig. 7. Results of encryption by using 3 pixels x 3 pixelsblocks. (a) Original image. (b) Encrypte d image usingBlowfish. (c) Transformed image. (d) Encrypted imageusing transformed followed by the Blowfish algorithm.
The results ofthis case are summarized in Table IV and Fig. 8.
Table IV Results of Correlation and Entropy values of Case 3.
Measurement A B C D
Horizontal 0.9325 0.0346 0.74 69 0.0245
c Vertical 0.9362 0.1073 0.7497 0.0067.S-ee
Diagonal 0.9186 0.0321 0.5881 0~t:0 OppositeU 0.9156 0.0337 0.5877 0.0056
Diagonal
Average 0.9257 0.05 19 0.66 81 0.0092
Entropy value 2.431 2.43 05 4.799 2.4305
(b)
Fig. 8. Correlation and entropy values of Case 3. (a)Correlat ions between directions. (b ) Average correlation and
entropy.
(a) (b)
(c) (d)411
IAENG International Journal of Computer Science, 35:1, IJCS_35_1_03
1.10.06 • • •_.. ._--. 0.90.05 .-- ..
j,<I>0.7 j,OJ
~ 0.04j,c 0.03 - 0.5 .
<I>o 0.02
j, 0 .3Qjn, 0.01 --. j, - - - - - - - - - -- 0.1 l- I -- Ij,
0 -0.1 - -
30 x 30 60 x 60 100 x100 30 x 30 60 x 60 100 x100
Nurrber of blocks Nurrber of blocks
-- ...- A • 6 ~ C ~- - - I:)
• B j, D(a)
(a)6
t t•6 5 -.....- - - j,
<I> 5 j, -- -+---- - -- ---. 4
OJ ..-
~ 4 3c3<I> 2 - • • •o
Qj 2n, 1 1
0 30 x 30 60 x 60 100 x100
30 x 30 60 x60 100 x1 00 Nurrber of blocks
Nurrber of blocks ...-+-A • B j, C j, 0
I • B j, 0_ 0_'_"'_- -- -- __.. •
(b)Fig. 9 Average correlation and entropy for image Bandimage D. (a) Correlation. (b) Entropy
As shown in Fig. 9 above, the three different cases showedthat the proposed algorithm resulted in lower correlation andhigher entropy than the Blowfish algorithm alone . The bestperformance occurs for the smallest image block size .
V. RESULTS AND DISCUSSION
The above cases show that using the proposed algorithmfollowed by the Blowfish algorithm resulted in a lowercorrelation and a higher entropy compared to using theBlowfish alone. Dividing the image into a larger number ofblocks made the performance even better. The resultsshowed that the correlation was reduced even further and theentropy was increased as the number ofblocks is increased.
Table V summarizes the results of the different cases , whileFig. 10 shows the effect of block size on the correlation andentropy value.
Table V Results of Correlation and Entropy values of image(a) in Fig. 3.
Measurement NumberA B C Dof blocks
Average30 x 30 0.7952 0.023460 x 60 0.9257 0.0519 0.6681 0.0092correlation100 -roo 0.5211 0.005630 x 30 5.2305
Entropy 60 x 60 2.4305 4.799 2.4305 5.4737100 - roo 5.5281
(b)Fig. 10. Correlation and entropy values ver sus the number ofblocks. (a) Correlation. (b) Entropy.
VI. COMPARISON BETWEEN THE PROPOSEDALGORITHM AND OTHER ALGORITHMS
In this section, two experiments were carried out usingthe image of Fig. 11. In the first experiment, the image wasdivided into different blocks that are shuffled according tothe proposed algorithm and then followed by one of fourcommonly used encryption algorithms; Blowfish, Two~sh,RijnDael , or RC4. The se algorithms are commerciallyavailable, so we applied them on the ciphered image thatresulted from applying the proposed algorithm on ~he
different block sizes of the original image. The correlationand entropy of each one was compared with applying t~ecorresponding algorithm alone. The results of thisexperiment are shown in Table VI and Fig. 12. .Fig. 12 shows that using the proposed algorithm along WIththe other algorithms resulted in a better performancecompared to using the other algorithms alone. .
In the second experiment, commonly used algorithms 10
literature were used to compare with the proposedalgorithm. Since those algorithms are not available for .us,we compared the ciphered images generated by applylOgthese algorithms with the ciphered images generated byapplying the proposed algorithm on the original image. Thecorrelation and entropy of the ciphered images w~re
recorded for each algorithm. The results of this applicatlonare shown in Table VII and plotted in Fig. 13.
412
IAENG International Journal of Computer Science, 35:1, IJCS_35_1_03
Fig. 11. Image used to compare the proposed algorithm with the commonly used ones in industry and literature
Table VI Correlation and Entropy when Appling. (a) Commercially available algorithms alone. (b) Commercially availablealgorithms preceded by the proposed algorithm when applied to the image of Fig. II
Commercially available algorithms aloneSystem Correlation Entropy
BLOWFISH(448) 0.0189 5.2703TWOFISH (256) 0.0054 5.5437RijnDael (AES256) 0.0051 5.5436RC4(2048) 0.0038 5.5437
(a)
Commercially available algorithms preceded by the proposed algorithm
System Number of blocks Correlation Entropy30x30 0.0063 5.4402
BLOWFISH(448)60x60 0.0049 5.5286lOOxlOO 0.0044 5.5407300x300 0.0028 5.543830x30 0.0026 5.5437
TWOFISH (256) 60x60 0.0040 5.5439ioo- 100 0.0041 5.5438300x300 0.0029 5.543830x30 0.0034 5.5440
RijnDael (AES256)60x60 0.0024 5.5439lOOxlOO 0.0049 5.5438300x300 0.0016 5.543930x30 0.0024 5.5438
RC4(2048)60x60 0.0026 5.5437lOOxlOO 0.0034 5.5439300x300 0.0034 5.5438
(b)
0.040.035
0.030.025
0.02 • • • •0.015
0.010.005 • • • • I I • 1 • • •
0 • • • • •0 0 0 0 0 0 0 0 0 0 0 0
'" <0 0 0 '" <0 0 0 '" <0 0 0x x ~ '" x x ~ '" x x ~ '"0 0 x x 0 0 x x 0 0 x x
'" <0 0 0 '" <0 0 0 '" <0 8 80 0 0 0~ '" ~ '" ~ '"
• • • •
0 0 0 0 0 0 8 0 0 0 0 0 0 0 0 0
'" <0 0 0 '" <0 0 '" <0 0 0 '" <0 0 0x x X '" x x X '" x x ~ '" x x X '"0 0 x 0 0 x 0 0 x x 0 0 x
'" <0 0 0 '" <0 0 0 '" <0 0 0 '" <0 0 00 0 ~ 0 0 0 0 0~ '" '" ~ '" ~ '"
BLOWFISH lWOFISH RijnDael RC4
• • • • • • • • • • • • • • •5.6
5.5•5.4
5.3
5.2
5.1
5
RC4
I I • •
RijnDaeIBLOWFISH 1WOFISH
• Other systems • Proposed technique • Other system; • A"oposed technique
(a) (b)Fig. 12. Correlation and entropy values of the algorithms shown in Table VI. (a) Correlation. (b) Entropy.
413
IAENG International Journal of Computer Science, 35:1, IJCS_35_1_03
Table VII Correlation and Entropy values ofcommonly used Encryption Algorithms in Literature when applied to the image ofFig. II.
Image encryption algorithmMeasurement
NumberCorrelation Entropy
I Proposed technique 30 x 30 0.0063 5.44022 Proposed technique 60 x 60 0.0049 5.52863 Proposed technique 100 x 100 0.0044 5.54074 Block Permutation 8x8 0.7977 1.9216
5 Pixel Permutation 0.741 8 1.9017
6 Bit Permutation 0.0479 1.97697 [Block, bit, pixel] Combination 0.0812 2.0631
8 3D Jigsaw transform 0.6096 2.3735
9 Arnold cat map 0.2882 1.9208-,... _-- -.- '. -'~. ., ". . ' ~...-
j,- .."•....~....,..,-•....~
- j!~O Chen's O~2 569 :,',. 3;;·140
I·i
l"" ,.....O. g07 ~ I 2.0635
;
i I 1 A naive selective encryptiun (3 oits eliCfyptea j iI 12 A naive selective encryption (7 bits encrypted) 0.1037 2.0182
" "
".,.
• • ":-_ ·· · _ -- - - -- -' 1 - .- --- - - - -_ . " ' - . - - " ' "
2 3 4 5 6 7 8 9 10 11 12 2 3 4 5 6 7 8 9 10 11 12
0.9
0.80.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
-0.1
• •"
6
5
4
3
2 ·
o
. .. •
... . . ••.. .. •
~) ~)
Fig . 13. Correlation and entropy values of the algorithms shown in Table VII. (a) Correlation. (b) Entropy.
Fig. 12 and Fig. 13 show that the proposed algorithm resultedin lower correlation and higher entropy and thus in a betterencrypted image than all other algorithms tested in this paper.
VII. CONCLUSION
In this paper a simple and strong method has beenproposed for image security using a combination of blockbased image transformation and encryption techniques. Thecase s showed that the correlation was decreased when theproposed algorithm was applied to them before the Blowfishalgorithm. Experimental results of the proposed techniqueshowed that an inverse relationship exi sts between numberof blocks and correlation, and a direct relationship betweennumber of blocks and entropy. When compared to manycommonly used algorithms, the proposed algorithm resultedin the best performance; the lowest correlation and thehighest entropy.
REFERENCES
[ I J W. Lee, T. Chen and C. Chieh Lee, "Improvement of anencryption scheme for binary images," Pakistan
Journal of Information and Technology . Vol. 2,no. 2, 2003 , pp. 191-200. http ://www.an sin et. org/
[2J M. V. Droogenbroech, R. Benedett, "Techniques for aselective encryption of uncompressed and compressedimages," In ACIVS '02, Ghent, Belgium. Proceedingsof Advanced Concepts for Intelligent Vision Systems,2002.
[3J S. Changgui, B. Bharat, "An efficient MPEG videoencryption algorithm," Proceedings of thesymposium on reliable distributed systems, IEEEcomputer society Press, 1998, pp. 381-386.
[4J S. Fong, P.B. Ray, and S. Singh, "Improving thelightweight video encryption algorithm," proceedingof iasted international conference, single processing,pattern recognition and application, 2002, pp. 2528 .
[5J S. P. Nana'vati., P. K. panigrahi. "Wavelets:applications toimage compression- I,". joined of the scientific andengineering computing, vol. 9, no . 3, 2004, pp. 4- 10.
[6] c. Ratael, gonzales, e. Richard, and woods, "Digitalimage processing," 2nd ed, Prentice hall, 2002.
[7] AL. Vitali, A. Borneo, M. Fumagalli and R . Rinaldo,"Video over IP using standard-compatible multipledescription coding," Journal of Zhejiang
414
IAENG International Journal of Computer Science, 35:1, IJCS_35_1_03
University- Science A, vol. 7, no. 5 ,2006, pp. 668676 .
[8] H. EI-din. H. Ahmed , H. M. Kalash, and O. S. FaragAllah, "Encryption quality analysis of the RC5 blockcipher algorithm for digital images," MenoufiaUniversity, Department of Computer Science andEngineering, Faculty oj Electronic Engineering,Menouf-32952, Egypt, 2006.
[9] Li. Shujun, X. Zheng "Cryptanalysis of a chaotic imageencryption method," Inst. of Image Process. Xi'anJiaotong Univ., Shaanxi , This paper appears in: Circuitsand Systems, ISCAS 2002 . IEEE InternationalSymposium on Publication Date: 2002, Vol. 2,2002 , page(s):708,71 I.
[10] S.S. Maniccam, N.G. Bourbakis , "Image and videoencryption using SCAN patterns," Journal ofPattern Recognition Society , vol. 37, no. 4, pp.725737 , 2004.
[II] A. Mitra , , Y V. Subba Rao, and S. R. M. Prasnna, "Anew image encryption approach using combinationalpermutation techniques," Journal oj computer Science,vol. I, no. I, p.127, 2006, Available :http://www.enfo rmatika .org
[12] G. Zhi-Hong, H. Fangjun, and G .Wenjie ,"C h ao s- based image en cryption algorithm,"Department oj Electrical and computer Engineering,University of Waterloo, ON N2L 3G I, Canada.Published by: Elsevier, 2005 , pp. 153-157.
[13] A. Sinha, K. Singh, "Image encryption byusing fractional Fourier transform andJigsaw transform in image bit planes ,"Source: optical engineering. spie-int society opticalengineering, vol. 44, no. 5 , 2005, pp.15-18.
[14] Li. Shujun, Li. Chengqing, C. Guanrong, Fellow.,IEEE., Dan Zhang., and Nikolaos,G., BourbakisFellow., IEEE. "A general cryptanalysis ofpermutation-only multimedia encryptionalgorithms,"2004,http://eprint.iacr. Org/2004/374.pdf
[15] I. Ozturk, I.Sogukpinar , "Analysis andcomparison of image encryption algorithm,"Journal oj transactions on engineering, computingand technology December, vol. 3, 2004, p.38 .
http : //www.enfonnatika.org![16] A . Sinha, K. Singh, "A technique for image
encryption using digital signature ," Source:Optics Communications , vol.218, no . 4, 2003,pp.22 9-234. http: / /www. els e vi er .com/
[17] S. S . Man icc am . , G . N i k 0 Ia 0 s , an dBourbakis, "Lossless image compression andencryption using SCAN ," Journal of Pattern Recognition.vol. 34, no. 6: , 2001 , pp.1229- 1245.
[18] M. Sonka, V. Hlavac . and R. Boyle, "Digital imageprocessing," in : image Processing, Analysis, andMachine Vision, 1998,2nd ed. htt[i:l/\\ ·\\'lI '.p\l·s.COIll
[19] D. Feldman, "A brief introduction to: informationtheory, excess entropy and computational mechanics,"college of the atlantic 105 eden street, bar harbor, me04609,2002, http:/ /hornacek.coa.edu/
Mohammad. A. B. Younes received BSc in computer science from YarmoukUnivers ity in Irbid Jordan, MSc from Sudan University of science and
technology in 1986 and 2003 respectively and currently enrolled in the PhDprogram in computer science in the Universiti Sains Malaysia.
Dr. Aman Janran received BSc in computer science, Master in AI, PhD inSoftware Engineering from Universiti Sains Malaysia, penang in 1993. 1996and 2002 respecti vely and currently Lecturer in School of ComputerSciences in Universiti Sains Malaysia. penang.
415